Rainfall Sensor Based On Coarse And Fine Segmentation Method Of Raindrop

Vehicle rainfall sensor is the core part of an automatic wiper.It has great significance for improving driving safety and intelligence of vehicle.Although infrared and capacitive rainfall sensors which are the mainstream of the market have already had mature products,they all have inherent shortcomings that are difficult to overcome.The research of the visual rainfall sensor is not mature,and it is difficult to take into account the two aspects of cost and performance.To solve the above problems,a rain sensor based on coarse and fine raindrop segmentation is studied under general imaging conditions.The improved background blur gradient method is used to make the raindrop coarse segmentation.The raindrop fine segmentation is realized by the raindrop shape feature and the luminance partition method.The algorithm is simple and fast,which ensures the detection precision and processing speed of the detection algorithm under the low cost hardware condition.The main contents of this paper are as follows:In view of the gray and gradient characteristics of the raindrop,the raindrop coarse segmentation algorithm based on the improved background blur gradient method is designed by using the characteristic that windshield raindrops will keep static in short time.First,average several successive frame images,and then the image background will be further blurred by the set operation between smooth and average images.After detecting edges with the Sobel operator,the edge that is not belong to raindrops will eliminated by the strongest gradient method,and the coarse segmentation of the raindrops is completed.This method can get nearly 80% correct detection rate,which can guarantee lower false positive rate and missed detection rate in some scenarios.In view of the missed detection in the bright area,the power law transformation is used to enhance the gradient characteristics of the raindrops in bright area,and the Sobel operator is used to segment the bright area raindrop after the noise reduction.The test results of several scenes show that the proposed method effectively improves the detection of bright area leakage.For raindrop coarse segmentation results,there are some problems of false detected.Raindrops are segmented precisely according to raindrop shape characteristics.By eliminating the connected domain with large area and low roundness in the coarse segmentation results,the false detection which is not consistent with the raindrop feature can be effectively eliminated.Finally,the algorithm is encapsulated in the software,and the rainfall information is calculated to achieve the determination of rainfall.The results of experiment show that the algorithm has nearly 91% positive detection rate,the false detection rate and the missed detection rate are 9% and 4.4% respectively.The results of rainfall determination are in line with the driver's senses.Compared with the similar method,the algorithm can work in a general imaging condition,and maintain a high detection precision,the computation speed and the adaptability to the complex environment are also improved significantly.